Verify and expand tracing instrumentation to cover all LLM backends consistently

[psschwei]

Description:
Verify and expand tracing instrumentation to cover all LLM backends consistently.

Detailed Requirements:

1. Audit current backend instrumentation:
   • OpenAI: Well instrumented (reference implementation)
   • Ollama: Verify/add instrumentation
   • HuggingFace: Verify/add instrumentation
   • WatsonX: Verify/add instrumentation
   • LiteLLM: Verify/add instrumentation
   • vLLM: Verify/add instrumentation
2. Ensure each backend has:
   • instrument_generate_from_context() for chat
   • instrument_generate_from_raw() for completion
   • Token usage recording
   • Error handling with semantic types
3. Add backend-specific attributes where relevant
4. Create instrumentation checklist/tests

Files to Modify:

• mellea/backends/ollama.py
• mellea/backends/huggingface.py
• mellea/backends/watsonx.py
• mellea/backends/litellm.py
• mellea/backends/vllm.py

Backend Instrumentation Checklist:

Backend	generate_from_context	generate_from_raw	Token Recording	Error Handling
OpenAI	Yes	Yes	Yes	Yes
Ollama	?	?	?	?
HuggingFace	?	?	Limited	?
WatsonX	?	?	Yes	?
LiteLLM	?	?	Yes	?
vLLM	?	?	?	?

Acceptance Criteria:

• All backends have consistent instrumentation
• Token recording works for all backends that support it
• Errors handled consistently across backends
• Instrumentation patterns documented

[ajbozarth]

Closing as part of the tracing epic (#444) re-plan. This sub-issue predates the plugin/hooks pattern (PR #653) and the GenerationMetadata partition (PR #908); its spec has drifted.

After the Phase 1 plugin refactor, a backend coverage audit becomes a test-parity check rather than a survey of duplicated inline instrumentation.

Replacement: #1054 — backend coverage parity audit post-refactor.

See #444 for the full refreshed plan.